HEK293T cells (ATCC: #CRL-3216) were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS), 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine. Wild-type HCT116 (ATCC: #CCL-247) and LSR-deficient HCT116/ΔLSR_CRISPR cells (Hemmasi et al., 2015) were cultivated in DMEM supplemented with 10% (v/v) FCS, 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine. HeLa cells were grown in Minimum Essential Medium (MEM) supplemented with 10% (v/v) FCS, 1% (v/v) penicillin/streptomycin, 2 mM L-glutamine, and 1 mM sodium pyruvate. A431 cells were grown in Roswell Park Memorial Institute (RPMI) 1640 Medium supplemented with 10% (v/v) FCS, 1% (v/v) penicillin/streptomycin, and 2 mM L-glutamine.

For HCT116 cells, Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific, Waltham, United States) and for 293T-GFP cells, TransIT-LT1 Transfection Reagent (Mirus Bio LLC, Madison, United States) were used for transfection, according to the manufacturer’s protocols.

Cell viability measurements with cultured cells were performed with the CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay (MTS) from Promega (Madison, United States), according to the manufacturer’s protocol. Briefly, cells were seeded in 96-well plates and the next day the medium (100 µL) was exchanged with medium (100 µL) containing the respective toxin components at indicated concentrations. Mock- and DMSO (10% (v/v))-treated cells served as positive and negative control for cell viability, respectively. Following an incubation period of 24 h at 37°C, MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) was added to the medium of each well and the absorption was measured at 490 nm using a microtiter plate reader.

Monoclonal cell populations were obtained by limited dilution cloning. Briefly, suspensions of trypsinized cells were adjusted with growth medium to a cell density of 5 cells per ml and transferred to 96-well plates in 100 µL portions (corresponding to 0.5 cells per well). An inverted microscope Axiovert ProgRess C10 plus was used to detect wells that contained only one defined cell clone after 7 days of growth. Single cell colonies were then removed by trypsinization and transferred to culture dishes for expansion.

GFP expressing HEK293T cells were generated with third generation lentiviral particles (Dull et al., 1998) as described previously (Koepke et al., 2020). In short, third generation lentiviruses were produced using pCSC-SP-PW-GFP and helper plasmids and then used to transduce HEK293T cells. Three days post transduction, the cells were sorted using the BD FACSAria II. Cells with moderate GFP MFIs were pooled and cultivated at 37°C with 5% (v/v) CO₂ under humidified conditions. The pCSC-SP-PW-GFP plasmid (aka pBOB-eGFP) was kindly provided by Inder Verma (Addgene, #12337 (Marr et al., 2004)).

Recombinantly expressed and purified toxins or toxin components, such as His-TccC3hvr, C2I, C2II, CDTa, CDTb, PA (PA83) and the receptor-binding domain (RBD) of CDTb, are described elsewhere (Barth et al., 2000; Tonello et al., 2004; Lang et al., 2010; Schwan et al., 2011; Papatheodorou et al., 2013). The RBD of CDTb was used as a glutathione S-transferase (GST) fusion protein. The binding components of CDT, C2 and anthrax toxin, namely CDTb, C2II and PA83, respectively, were used as protease-activated proteins, whereas protease-activated PA83 is denoted as PA63 (Singh et al., 1989; Barth et al., 2000; Blöcker et al., 2001; Sterthoff et al., 2010). Unnicked diphtheria toxin (DT) from Corynebacterium diphtheriae was purchased from Sigma-Aldrich (order number: D0564). His-TccC3hvr, both CDT components (CDTa and CDTb) and the RBD of CDTb were provided by the laboratory of Klaus Aktories (University of Freiburg, Germany). The PA variants mPA83 and mPA83-EGF as well as LF_N-DTA were provided by R. John Collier (Harvard Medical School, Boston, United States).

Subconfluent monolayers of cultured cells were intoxicated by direct addition of toxins or toxin components into the growth medium. If required, toxin dilutions were performed with growth medium or phosphate-buffered saline. If not otherwise stated, the final concentrations of the toxins during the intoxication were as follows: CDTa/CDTb (5 nM/1 nM), C2I/C2II (1 nM, 1.67 nM), PA63/His-TccC3hvr (10 nM/50 nM), DT (10 nM), and the combinations PA83, PA63, mPA83 or mPA83-EGF (20 nM) plus LF_N-DTA (10 nM), respectively. The intoxication of cells was analyzed by microscopic evaluation of the cell morphology, by measuring cell viability or by immunodetection of target substrate modification.

For the recombinant production of His-Cas9, Escherichia coli XJb (DE3) Autolysis strain (T3051, Zymo Research Europe, Freiburg, Germany) was transformed with Addgene plasmid #53261 (pET28a/Cas9-Cys (Ramakrishna et al., 2014)). Briefly, transformants were grown in liquid LB medium at 37°C until an OD_{600 nm} of 0.6–0.8 was reached. Then, 0.5 mM IPTG and 3 mM arabinose were added to the medium for His-Cas9 and lambda lysozyme expression, respectively, followed by incubation for 4 h at 37°C under shaking. After harvesting of the cells by centrifugation, cells were resuspended in buffer, containing 50 mM NaH₂PO₄ (pH 8), 300 mM NaCl, 5 mM imidazole, 0.03% (v/v) Triton X-100, and stored at −20°C. A cell lysate for subsequent protein purification was generated by thawing the cell pellet, followed by addition of 1 mM PMSF and 1 mM DTT to the cell suspension. Next, cell debris was removed by centrifugation (30,000 g, 30 min, 4°C) and the supernatant filtered through a 0.45 and 0.2 µm sterile filter. Finally, the filtrate was subjected to PureCube 1-step batch Midi Plus Columns (Cube Biotech, Monheim, Germany) for purification of His-Cas9 by nickel affinity chromatography. Washing and elution buffers were 50 mM NaH₂PO₄ (pH 8), 300 mM NaCl, 20 mM imidazole, 0.03% (v/v) Triton X-100 and 50 mM NaH₂PO₄ (pH 8), 300 mM NaCl, 250 mM imidazole, respectively. Eventually, a buffer exchange of the eluate to phosphate-buffered saline (PBS) was performed with Vivaspin 20 columns (Sigma-Aldrich Chemie GmbH, Munich, Germany).

LF_N-Cas9 was expressed and purified essentially as described for His-Cas9 above. However, the initial cloning of the LF_N-Cas9 fusion construct was necessary. The Addgene plasmid #62933 (pET-Cas9-NLS-6xHis (Zuris et al., 2015)) served as backbone for insertion of an LF_N-encoding fragment upstream of the Cas9 sequence by seamless assembly (NEBuilder HiFi DNA Assembly Cloning Kit; #E5520; New England BioLabs GmbH, Frankfurt am Main, Germany). The LF_N-encoding fragment was generated by PCR and with Addgene plasmid #11075 (pET-15b LFN-DTA (Milne et al., 1995)) as template. The NEBuilder Assembly Tool (New England BioLabs GmbH, Frankfurt am Main, Germany) was used to design primers for the seamless cloning procedure.

Purified His- and LF_N-Cas9 proteins were tested for dsDNA cleavage activity by performing an in vitro Cas9 cleavage assay with the CRISPRcraft S. p. Cas9 Control Kit (Lucigen, Middleton, United States), which includes a substrate DNA, a substrate DNA-specific gRNA and a set of DNA primers. For the in vitro cleavage assay, ∼300 ng of a 1083 bp DNA fragment (generated by PCR and by using the supplied substrate DNA as template) was incubated for 10 min together with the gRNA and 20 nM or 40 nM His-Cas9- and LF_N-Cas9, respectively, and cleavage products were analyzed by agarose gel electrophoresis. The gRNA-directed cleavage of the 1083 bp fragment results into two fragments with the size of 800 and 283 bp, respectively. Commercial Cas9 protein (EnGen^® Spy Cas9 NLS from New England Biolabs (NEB); M0646T) was used as positive control for in vitro DNA cleavage.

For the detection of His- and LF_N-Cas9 via immunoblotting and enhanced chemiluminescence (ECL) imaging, a mouse anti-His (AD1.1.10) antibody (Santa Cruz Biotechnology, Dallas, United States) as primary antibody in combination with mouse IgGκ light chain binding protein coupled to horseradish peroxidase (m-IgGκ BP-HRP) was used.

Whole-cell lysates were generated directly in well plates or culture dishes by washing cell monolayers first with PBS and then by resuspending them in Laemmli buffer. Prior to SDS-PAGE followed by immunoblotting, lysate samples were heated at 95°C for 5 min. For the detection of LSR, Hsp90 and GAPDH, commercial primary antibodies (rabbit anti-LSR (X-25); mouse anti-Hsp90 α/β (F-8); mouse anti-GAPDH (G-9)) and secondary antibodies (m-IgGκ BP-HRP and goat anti-rabbit IgG-HRP) were used. All antibodies were purchased from Santa Cruz Biotechnology, Dallas, United States.

An inverted Axiovert 40 CFL microscope (Carl Zeiss Microscopy, Jena, Germany) equipped with a ProgRess C10 plus camera (Jenoptik, Jena, Germany) was used in this study for light microscopic analyses, such as monitoring cell intoxication, identification of growing cell clones in wells and cell confluence estimation, respectively.

First, genomic DNA was isolated from 1 × 10⁶ cells using the MyTaq Extract-PCR Kit (Bioline GmbH, Luckenwalde, Germany) or innuPREP DNA Mini Kit (Analytik Jena, Jena, Germany) following the manufacturer’s recommendations. Next, a 403 bp DNA fragment within the lsr gene that contains the LSR-specific gRNA binding site was amplified by PCR with oligonucleotides 5′-GTC​CAA​CCC​CTA​CCA​CGT​GGT​G-3′ and 5′- GCT​TTC​AGA​TGG​GGA​CTC​CAG​G-3′ and by using genomic DNA as template (Hemmasi et al., 2015). Eventually, the PCR product was purified with the my-Budget Double Pure Kit (BioBudget Technologies GmbH, Krefeld, Germany) and subjected to DNA sequencing (Eurofins Genomics Europe Sequencing GmbH, Konstanz, Germany). Indels in the sequenced DNA fragments were identified by performing sequence alignments with the lsr reference sequence.

The sequential ADP-ribosylation assay served to estimate in vitro the ADP-ribosylation status of cellular actin after intoxication of cells by the actin ADP-ribosylating toxin CDT (Kaiser et al., 2011). First, intoxicated cells (and mock-treated cells as negative controls) were washed twice with PBS and frozen at −20°C. Then, cell lysis was induced by thawing, after which the lysed cells were collected with a cell scraper and resuspended in buffer containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl₂, 1 mM EDTA, 1 mM DTT, and cOmplete protease Inhibitor Cocktail (Roche, Rotkreuz, Switzerland). Cell lysate (20 µL) was incubated for 30 min at 4°C together with purified CDTa (100 ng) and 250 pmol biotinylated NAD⁺ (Trevigen, Gaithersburg, United States), prior to addition of Laemmli buffer and heating for 10 min at 95°C to stop the ADP-ribosylation reaction. Eventually, in vitro sequentially ADP-ribosylated (and thus biotinylated) actin was visualized by Western Blotting and ECL and by the use of peroxidase-conjugated streptavidin (Merck, Darmstadt, Germany).

For the analysis of cell surface binding, purified proteins were fluorescently labelled with DyLight 488 NHS Ester (Thermo Fisher Scientific, Waltheim, United States) following manufacturer’s instructions and as described earlier (Papatheodorou et al., 2013). Zebra Spin Desalting Columns (Thermo Fisher Scientific, Waltham, United States) were used to remove excess fluorescent dye after the labelling reaction. Labelling efficiency and protein concentration were determined by performing spectrophotometric analyses on a NanoDrop OneC spectrophotometer (Thermo Fisher Scientific, Waltham, United States). Next, cultured cell monolayers were detached by incubation for 15–30 min at 37°C with PBS containing 25 mM EDTA (and without trypsin) to preserve proteins at the cell surface. The detached cells were resuspended in growth medium and the cell density of the cell suspension adjusted to 1 × 10⁶ cells per ml. Then, DyLight-labelled protein at indicated amounts was incubated with 2 × 10⁵ cells (200 μL cell suspension) for 30 min on ice to prevent cellular uptake of cell surface-bound proteins. Finally, cells were washed up to 5 times by centrifugation (500 g, 5 min, 4°C) and resuspension in PBS to remove any non-bound DyLight-labelled protein, prior to flow cytometric analysis of the cell suspensions using the BD FACSCelesta flow cytometer run by the FACSDiva software (BD Biosciences, San Jose, United States). The results of the flow cytometric measurements (typically 1 × 10⁴ cells per sample) were further analyzed using Flowing software version 2.5.1 (University of Turku, Finland).

The sorting of HEK293T cells expressing moderate GFP levels was done using the BD FACSAria III. To measure GFP fluorescence of cells applied to the CRISPA method the Beckman-Coulter CytoFLEX was used.

Guide RNA (gRNA) expression plasmids including the protospacer sequences 5′- GGA​CAG​CGT​GCG​CAC​CGT​CA-3 for LSR and 5′-GTG​AAC​CGC​ATC​GAG​CTG​AA-3′ for GFP were used in this study that are described elsewhere (Hemmasi et al., 2015; Koepke et al., 2020).

293T-GFP cells were cultivated in a 24-well to half-confluency, then transfected with a plasmid encoding a GFP-specific gRNA, followed by medium renewal after 6 h. The next day, the cells were incubated for 24 h with 10 nM PA63 together with 100 nM His- and LF_N-Cas9, respectively, or a commercially available His-Cas9 protein (EnGen^® Spy Cas9 NLS from New England Biolabs (NEB); M0646T). Cells were then harvested for T7 endonuclease I (T7EI) mismatch detection using the GeneArt^® Genomic Cleavage Detection Kit (Thermo Scientific). In short, total genomic DNA was extracted from harvested cells and the locus, which is targeted by the GFP-specific gRNA was amplified by PCR using the primers 5′-ACA​GCT​CGT​CCA​TGC-’3 (GFP-rev) and 5′-TGC​TTC​AGC​CGC​TAC​C-’3 (GFP-fwd). Subsequently, the PCR product was heat-denatured and reannealed to generate mismatch strands. These mismatches were then detected and cleaved by the Detection Enzyme (T7EI) and the resulting products were separated by agarose gel electrophoresis and visualized by gel imaging (Gel Doc XR + Gel Documentation System; Biorad). DNA band intensity was analyzed with ImageJ (Schneider et al., 2012) and the cleavage efficiency calculated essentially as described in the GeneArt^® Genomic Cleavage Detection Kit manual.

Bar diagrams were generated with mean values calculated from triplicates or from three samples performed in parallel and with error bars representing standard deviation (SD). Significance differences were calculated with Microsoft Excel (Student´s) t-test. Resulting p values were indicated by asterisks as follows: *p < 0.05, **p < 0.01, ***p < 0.001.

Here, we describe a novel modular PA-based Cas9 delivery concept, denoted as CRISPA, that can be subdivided into three main processes. The first is the transfection of a plasmid encoding the target guide RNA into cells. The second step is the delivery of Cas9 into cells via PA. As a consequence of both processes, a third process is initiated, the Cas9-dependent cleavage of the target gene, eventually resulting in the knockout of the respective gene (Figure 1A).

The CRISPA method requires purified recombinant Cas9 protein, either fused to an N-terminal His6-tag (denoted as His-Cas9) or to the N-terminus of the Bacillus anthracis lethal factor (denoted as LF_N-Cas9). Suitable Cas9 constructs must also possess a nuclear localization signal (NLS), in order to reach the nucleus of target cells. As His-Cas9, we used a recently described Cas9 construct consisting of an N-terminal His6-tag and Cas9 followed by an HA tag, an NLS and a single cysteine residue (Ramakrishna et al., 2014). As LF_N-Cas9, a Cas9 construct consisting of Cas9 followed by an NLS and a His6-tag was used, where we fused the N-terminus of the lethal factor (LF_N) at its N-terminus by Gibson cloning (Zuris et al., 2015) (Figure 1B). Both Cas9 fusion proteins were expressed in Escherichia coli and purified as His-tagged proteins by nickel affinity chromatography. The purity of both Cas9 preparations (∼87% for His- and 82% for LF_N-Cas9) was determined by SDS-PAGE and Coomassie staining (Figure 1C). Both Cas9 preparations were almost as active as commercial Cas9 protein in an in vitro Cas9 cleavage assay (Figure 1D).

We have chosen the HCT116 cell line as a model cell line for our CRISPA approach, because we previously succeeded in knocking out a gene of interest via the conventional CRISPR/Cas9 approach in these cells (Hemmasi et al., 2015). To examine whether HCT116 cells are a suitable target for the CRISPA approach, we first determined whether PA (in particular PA63 was applied in this study) is able to promote translocation of foreign cargo proteins into HCT116 cells. Here, we used His-tagged TccC3hvr (actin ADP-ribosyltransferase domain of the Photorhabdus luminescens PTC3 toxin) as toxic cargo protein for PA, since it was reported that this combination is toxic to various mammalian cell lines, such as HeLa and Vero (Lang et al., 2010; Ernst et al., 2017). Beitzinger et al. utilized 3 μg/ml (∼50 nM) proteolytically-activated PA (PA63) for transport of a His- or LF_N-fused cargo protein into human umbilical vein endothelial cells (HUVECs) (Beitzinger et al., 2012). We tested whether 10 nM PA63 would already be sufficient for translocating His-TccC3hvr at three different concentrations (20, 50, 100 nM) into HCT116 cells.

Mock-treated HCT116 cells or HCT116 cells treated for up to 24 h with 100 nM His-TccC3hvr or 10 nM PA63 alone did not show any morphological signs of intoxication (Supplementary Figure S1A). However, HCT116 cells incubated with the combination of 10 nM PA63 plus 20, 50 or 100 nM His-TccC3hvr, displayed strong cell rounding, a typical hallmark of TccC3hvr intoxication due to its molecular action on the actin cytoskeleton, already after 6 h of incubation (Supplementary Figure S1B and S1C). Thus, 10 nM PA63 is sufficient for translocating His- or LF_N-fused cargo proteins into HCT116 cells.

These results indicated that HCT116 cells express host receptor(s) for PA and permit transport of its cargo protein.

We next set about knocking out LSR in HCT116 cells by applying the CRISPA method (Figure 2A). At first, HCT116 cells were cultivated in a 24-well plate to half-confluency (day 0), then transiently transfected with a plasmid encoding the LSR-specific gRNA and the medium renewed 6 h after transfection. At day 1, cells were treated for 24 h with 10 nM PA63 together with 100 nM His-Cas9 or LF_N-Cas9, respectively, to induce the Cas9-based knockout of LSR. It is of note that already at this stage (day 2), single cell clones for further analysis and identification of cell clones carrying the desired gene knockout may be isolated.

LSR is the host entry receptor for CDT. This enabled us to further select for LSR-deficient HCT116 cells simply by intoxicating the cells after CRISPA with CDT. To this end, cells were trypsinized at day 2 and one-tenth of the cells was re-seeded into a 6-well to obtain segregated colonies of clonal derivatives at the plastic bottom. Grown cell colonies were then incubated twice for 24 h with CDT at day 5 and day 6. At day 7, CDT-resistant cell clones could easily be discriminated microscopically from non-resistant cell clones that exhibited massive CDT-induced cell rounding. Only a few CDT-resistant cell clones were observed in each 6-well (typically up to 10), independent of whether His- or LF_N-Cas9 was used together with PA63. Notably, CDT-resistant cell clones were only detected when all components of the CRISPA approach (LSR-specific guide RNA-expressing plasmid, PA63 and His-/LF_N-Cas9) were used in combination, indicating that CDT-resistant cell clones did not arise from random mutations due to genetic instability. CDT-resistant cells were expanded and reselected again with CDT (days 7–15, Figure 2B). Eventually, single cell clones were obtained in 96-wells by limiting dilution and cell colonies were clearly visible after approximately 1 week of incubation (>day 22) (Figure 2B).

To test whether the CRISPA approach was successful, we analyzed a single LF_N-Cas9- (HCT116/ΔLSR_{CRISPA/LFN-Cas9}) and a single His-Cas9-derived cell clone (HCT116/ΔLSR_{CRISPA/His-Cas9}), respectively, via LSR immunoblotting. To this end, whole-cell lysates were generated from both CRISPA-derived LSR knockout cell clones and, for direct comparison, from wildtype HCT116 cells (HCT116/WT). Whole-cell lysates from LSR-deficient HCT116 cells generated using conventional CRISPR/Cas9 approach (denoted here as HCT116/ΔLSR_CRISPR) (Hemmasi et al., 2015) were used as negative controls for LSR immunodetection. LSR immunoblotting confirmed the lack of LSR expression in HCT116/ΔLSR_CRISPR, but also in HCT116/ΔLSR_{CRISPA/LFN-Cas9} and HCT116/ΔLSR_{CRISPA/His-Cas9} cells (Figure 3A).

Next, we analyzed whether indel mutations in our target gene LSR are present in both CRISPA-derived LSR knockout clones. For that purpose, genomic DNA was isolated from each cell clone and from HCT116/WT and HCT116/ΔLSR_CRISPR cells and a region including the guide RNA target site of the LSR gene was amplified by PCR and sequenced. In comparison to HCT116 wildtype cells, both the LF_N-Cas9- and the His-Cas9-derived cell clone exhibited a frameshift mutation (single-guanine insertion) directly within the Cas9 cleavage site of the LSR guide RNA target site. A single-thymine insertion was found in HCT116/ΔLSR_CRISPR cells at this position (Figure 3B). Both mutations lead to premature stop codons within the open reading frame (ORF) of LSR, resulting in a loss of expression of LSR.

We then examined whether CRISPA-derived LSR knockout cell clones became resistant towards CDT intoxication, either (I) via microscopic analysis of cell morphology or (II) by probing the ADP-ribosylation status of actin. Here, the Clostridium botulinum C2 toxin served as a control, as it enters the target cells independent of LSR (Papatheodorou et al., 2011). Consequently, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells were intoxicated for 24 h with CDT or C2 toxin, followed by microscopic analysis of the cell morphology. Wildtype and HCT116/ΔLSR_CRISPR cells served as controls in the intoxication assay. As expected, cell rounding induced by the cell cytoskeleton-disrupting activity of CDT was observed in wildtype HCT116 cells, while LSR-deficient HCT116/ΔLSR_CRISPR, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells were resistant to CDT (Figure 4A). All cells were readily intoxicated by C2 toxin (Figure 4A), indicating that general endocytic processes are still functional in these cells.

To substantiate our finding that HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells are resistant to CDT intoxication, we probed the ADP-ribosylation state of actin in lysates of CDT-treated and non-treated HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells using an in vitro actin-ADP-ribosylation assay. Importantly, the amount of actin amendable to sequential ADP-ribosylation upon CDT intoxication was only decreased in HCT116 wildtype but not in LSR-deficient HCT116/ΔLSR_CRISPR, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells (Figure 4B).

To verify lack of the LSR receptor, we demonstrated that CDT is no longer able to bind to HCT116/ΔLSR_CRISPR, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells. CDT is an AB-type binary toxin, where its B (binding/translocation) component (CDTb) mediates binding to LSR via a C-terminal receptor-binding domain (RBD) (Papatheodorou et al., 2013). Thus, the RBD of CDTb was labelled with the fluorescent dye DyLight 488 and its binding to suspension cells from wildtype and LSR-deficient HCT116/ΔLSR_CRISPR, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells at 4°C (no endocytic uptake) was analyzed by flow cytometry. Only wildtype HCT116 cells exhibited an increase in cell surface-bound fluorescence after incubation with the fluorescent RBD, confirming that HCT116/ΔLSR_CRISPR, HCT116/ΔLSR_{CRISPA/His-Cas9} and HCT116/ΔLSR_{CRISPA/LFN-Cas9} cells are lacking LSR expression at the cell surface (Figure 4C).

In summary, our results showed that the CRISPA method allowed knockout of a specific gene in a human cell line.

We next sought to determine if the CRISPA approach is applicable in other cell lines and for different target genes. For that purpose, we used a HEK293T cell line stably expressing green-fluorescent protein (GFP). In a previous study it was already shown that HEK293T cells are susceptible to the lethal toxin of B. anthracis (Chavarría-Smith and Vance, 2013). Here, the target gene for the CRISPA approach was the non-native, chromosomally integrated coding sequence of GFP. Thus, GFP-expressing 293T cells (293T-GFP cells) transiently expressing a GFP-targeting gRNA were incubated with PA63 in combination with His- or LF_N-Cas9, respectively. Cells were then incubated for 6 days, due to long half-life (∼26 h) of GFP in mammalian cells (Corish and Tyler-Smith, 1999) and GFP expression subsequently analyzed via flow cytometry. A clear decrease of the median fluorescence intensity (∼15–25%) of the 293T GFP cell population was observed, only if PA63 was added together with 80 nM His-Cas9 and LF_N-Cas9, respectively (Figure 5A). Single treatment with either PA63 or any Cas9 variant did not affect GFP expression. This confirms cell entry of both Cas9 variants and consequently knockout of GFP via the PA transporter.

In order to estimate the genome editing efficiency in CRISPA-targeted 293T-GFP cells, a commercial kit-based T7 endonuclease 1 (T7EI) mismatch detection assay was performed. 293T-GFP cells transiently expressing a GFP-targeting gRNA were treated with PA63 and His- and LF_N-Cas9. 293T-GFP cells co-transfected with a Cas9-expressing plasmid or treated with PA63 together with commercial Cas9 protein were used as controls. We observed genome editing efficiencies of 3.6/3.8% for His-Cas9, 2.4/2.6% for LF_N-Cas9, 3.8/5.7% for commercial Cas9 (NEB), 20.4/20.5% for transfected Cas9 plasmid, and 7.8/8.8% for the internal control of the commercial kit (Figure 5B).

These results confirm that CRISPA approach can be applied to different cell lines and native or non-native target genes.

The modular CRISPA approach has advantages over other non-viral and transient Cas9 delivery methods, because it offers the possibility for a cell-specific genome engineering. Previous studies have shown that PA can be engineered to bind to other host entry receptors. For instance, a mutated PA83 variant unable to bind to its native receptors due to mutations in the receptor-binding domain (termed mPA83), was fused to EGF (epidermal growth factor) and thus redirected to EGF receptors (EGFR), which are highly expressed in many cancer cells (Mechaly et al., 2012; Zahaf et al., 2017).

To confirm that mPA83-EGF selectively targets EGFR-positive cells, we used A431 and HeLa cells, which are characterized by high (A431 cells) and low (HeLa cells) levels of EGFR expression, respectively (Zhang et al., 2015). Both cell lines were intoxicated with 20 nM mPA83-EGF as the EGFR-specific transporter and 10 nM LF_N-DTA as the toxic cargo component. In LF_N-DTA, the N-terminus of the anthrax toxin lethal factor (LF_N) is fused to the enzyme portion of the diphtheria toxin (DTA), which ADP-ribosylates the eukaryotic elongation factor 2 (eEF2) and thereby inhibits protein synthesis (Milne et al., 1995). Measuring the cell viability after 24 h of intoxication revealed that A431 (Supplementary Figure S2; black bars), but not HeLa cells (Supplementary Figure S2; white bars), were intoxicated with mPA83-EGF plus LF_N-DTA. Importantly, a decrease in cell viability in both cell lines was obvious with 20% (v/v) DMSO or when cells were intoxicated with positive controls (10 nM native diphtheria toxin (DT) and the combinations of 20 nM PA83 or 20 nM PA63 plus 10 nM LF_N-DTA, respectively), but not with negative controls (the combination of 20 nM mPA83 alone plus 10 nM LF_N-DTA or the single toxin components) (Supplementary Figure S2).

These results demonstrate that cell specificity of the CRISPA approach can be achieved by the use of a mutated PA variant fused to a cell surface-specific ligand.